I am inspired to share my trajectory!
PhD - Postdoc - Baby 1 - Startup 1 - Baby 2 👼 - Startup 2 - Principal Investigator - Adjunct Professor - Baby 3 - Baby 4 - Assistant Professor. And you know what? That’s ok!
An academic career progression can be perfectly linear:
PhD - Postdoc - Assistant Professor.
Here’s mine:
PhD - Postdoc 1 - Postdoc 2 - Postdoc 3 - Instructor - Senior Postdoc - Scientist II - Scientific Director - Assistant Professor.
And you know what? That’s ok.
AMA ;)
Wow, I just got nominated to
#Emmys
as a creative director alongside celebrated director Patrick Clair for
#Westworld
main title!
More details about our GAN-induced "Dreams of Westworld" project:
Congrats to my team
@enisimsar
@omerkirbiyik
.
#Emmys2020
Happy to announce our collaboration with the celebrated director Patrick Clair () where we created an AI to dream Westworld. Included in the main title of WW this season! Read more in Syfy: Thanks to my team
@enisimsar
@omerkirbiyik
.
This pic made me realize what a blast the last 18 months have been. I welcomed Ada’s brothers Deniz & Derin. At the same time I started a lab mostly with undergrads where we published 11 papers at venues like ICCV, ECCV, WACV, CVPR and NeurIPS. 1/3
#AcademicTwitter
#AcademicMom
Living the nerd dream: I just gave a talk on Human-AI Collaboration while wearing an AI-generated dress, necklace, and perfume (and entered the stage with an AI-generated music! All courtesy of
@howtogaa
and collaborators! 🎉🤖
#congresofuturo2019
I am so excited that "How to Generate (Almost) Anything"
@howtogaa
is now officially a class at
@Bogazici_CmpE
where I am visiting this year! The focus is on Generative Models x Human-AI Collaboration. Some of the projects: carpets, men fashion, stained glass, tea, theater plays!
So proud of the senior thesis projects of the
@Bogazici_CmpE
students I worked with this semester! One project won the best thesis award and another (co-supervised with
@uskudarli
) took third place. Here are some examples of what they have accomplished this semester! ↘️
I am so shocked by the lack of depth in the
#ECCV24
reviews, instead of prepping for rebuttal I made this video from our stellar group of reviewers (volume up). Shout out to that last reviewer — you are the reason we still believe in academia.
@eccvconf
#reviewsystemisbroken
Happy to announce our collaboration with the celebrated director Patrick Clair () where we created an AI to dream Westworld. Included in the main title of WW this season! Read more in Syfy: Thanks to my team
@enisimsar
@omerkirbiyik
.
Proud advisor moment: all four of my PhD students have secured internships at top research labs in their first year (3 at Adobe Research and 1 at Amazon Research). Looking forward to seeing the exciting projects they will be contributing to!
The very first paper from my research group is just got accepted to
@ICCV_2021
! Congrats to my amazing team of students: Enis
@enisimsar
, Oguz
@okyksl
, and Gulperi!
Today is the day! Tune to FXX at 8PM EDT for
#Emmys
, where I am nominated for creative director for main title design (Westworld) alongside the genius director Patrick Clair ✨
#CVPR2024
was a blast! So proud of my PhD students Hidir, Tuna
@tunahansalih
, Yusuf
@yusuf_dalva
presenting 3 main and 2 workshop papers in the first year of their PhDs! I am also very proud of Kiymet
@akdemir_kiymet
presenting the first paper of her PhD work at
@iccc_conf
in
Two papers accepted to CVPR 2022 "Computer Vision for Fashion, Art, and Design" Workshop. Congrats to CATLAB undergrads Elif, Berkay, Merve, Umut and Kerem! Coming on Arxiv soon, but here is a teaser where we find top keyword-based directions in a model in an "unsupervised" way.
Initial
#styleGAN
results on the graffiti dataset we had for AI-generated graffiti piece back in summer for
@howtogaa
. Just can't wait to see how the results will come out with more training! For the earlier AI-generated graffiti work, see this:
Our paper "Graph2Pix: A Graph-Based Image to Image Translation Framework" is accepted to AIM 2021 Workshop at ICCV as a full paper. Congrats Dilara
@dilaragoekay
, Enis
@enisimsar
, Alper
@alperahmetoglu
, Efehan and Atif!
Just in time for London Fashion week, BBC interviewed me & my students at London College of Fashion about potential applications of AI & fashion. Stay tuned for their program on Saturday!
@FIALondon
@LCFLondon
@drinkmatt
Proud to announce that I am giving a new course called "AI and Fashion" at London College of Fashion this semester! We have an amazing group of students (all female!) who are super excited to work with AI!
@FIALondon
@LCFLondon
@drinkmatt
Dear Twitter: please meet with these new people I just 'generated'. These are the seed=132019 family. Reply to this tweet with a random number, and I will create a new AI-generated family for you!
#stylegan
#gan
#ai
#fun
Excited to share that our paper "ORACLE: Consistent Character Generation with LoRAs in Diffusion Models" has been accepted at ICCC'24! Congratulations to Kiymet
@akdemir_kiymet
on her first publication in the first year of her PhD!
@iccc_conf
#VirginiaTech
This reminded me of the time I enrolled in a fashion school after developing an interest in fashion at MIT. For a year, I spent entire weekends learning how to create garment patterns, design, sew, and more. A senior professor openly ridiculed me for "wasting my time," saying,
Stylebreeder
Exploring and Democratizing Artistic Styles through Text-to-Image Models
Text-to-image models are becoming increasingly popular, revolutionizing the landscape of digital art creation by enabling highly detailed and creative visual content generation. These
Three new pre-prints from GEM Lab!
GANtastic: GANs are known for their disentangled latent spaces & diffusion models are known for high-quality generation. Bring best of the both worlds and transfer GAN directions to diffusion!
@yusuf_dalva
@d_yesiltepe
This paper is also monumental for me since we pulled an allnighter for the
@ICCV_2021
deadline and I literally rushed to the hospital right after the submission to deliver my baby boy🤦♀️
Happy to announce that 2 papers are accepted to
#WACV2023
conference. One spotlight paper led by undergrads Zehranaz and Furkan, another led by CATLAB undergrad alumni Enis. Congrats to co-authors Alara, Umut and Gulperi as well!
@wacv_official
@Bogazici_CmpE
On the way to Seattle for
#CVPR2024
. DM me if you want to grab a coffee during the conference! Also check out 3 papers from GEM Lab ():
1) NoiseCLR (Oral), if you are into latent space:
2) RAVE (Highlight), if you are into video
Proud advisor moment, Yusuf
@yusuf_dalva
is giving an oral talk at
#CVPR2024
in the first year of his PhD! Check out NoiseCLR at and more work from GEM Lab at !
I am a mom of three and an assistant professor at Virginia Tech. The only way I can attend
@GHC
—the largest gathering of women and non-binary technologists— is by locking my kids in a hotel room with a stranger.
Happy to announce that our recent paper "MIDISpace: Finding Linear Directions in Latent Space for Music Generation" is accepted to ACM Conference on Creativity & Cognition. Congrats Meliksah and Alara!
@alexjc
@fhuszar
It is actually just
#styleGAN
! The algorithm works so well that (final results: ) we actually had to go to earlier iterations to find 'creative' designs (when the model was still 'learning' and making mistakes like weird color patches) :-)
More fun with
#styleGAN
. Apparently I will become a hairy man when I got old! (also kudos to my younger version for the nice nose job) Reply to this tweet w/your pic and I might do yours! 🥳
"StyleMC: Multi-Channel Based Fast Text-Guided Image Generation and Manipulation" is accepted to
#WACV2022
(). Congrats to Umut Kocasari (undergrad researcher!), Alara Dirik and Mert
@merttiftikci
. Coming on Arxiv soon!
Introducing NoiseCLR; an unsupervised approach to identify disentangled and interpretable directions in Stable Diffusion, without requiring any user-provided text prompts, or labeled data! First PhD work of my student
@yusuf_dalva
✨
Check out: .
I am so in love with the business cards I printed for
#CVPR24
! 6 members from GEM Lab will be in Seattle next week, check out the 3 papers we will present on our website:
Elif Sema Balcioglu, Berkay Doner and Merve Rabia Barin worked on one-shot discovery of interpretable directions in GANs by sketching and the results are amazing!
Long story short, I co-founded an AI & fashion company, taught the world's first AI-focused fashion course at the London College of Fashion () with course projects showcased at London Fashion Week (shout out to amazing
@drinkmatt
). 2/n
Contrast is all you need for high fidelity text-to-image diffusion models! First work from my PhD student
@tunahansalih
, in collaboration with Google (
@enisimsar
@fedassa
). Our method is training free, improves results in pre-trained SD and Imagen. Web: .
Exciting News! First research paper from my PhD, a collaboration between Virginia Tech and Google!
@PINguAR
@enisimsar
@fedassa
We introduce a novel approach to enhance the fidelity of text-to-image diffusion models.
Excited to share that our paper "MIDISpace: Finding Linear Directions in Latent Space for Music Generation" has received an honourable mention award at the ACM Creativity & Cognition Conference
@acm_cc
. Congrats Meliksah and Alara!
Cemre Efe Karakas and Eylul Yalcinkaya won the award for best senior thesis, in which we worked on editing GAN models to remove bias related to certain attributes such as gender or race.
@howtogaa
@Bogazici_CmpE
@iambatukoyuncu
Atif Yuksel and Sadullah Gultekin worked on a GAN-based men-fashion project and even featured some of the designs during their poster presentation! (needless to say I had to threaten them with an F grade if they don't wear the designs --including the skirt!)
I couldn't have found a better title ("Chaos Coordinator") to summarize my first year as an assistant professor (FYI: I scored this sign at Dollar General last week).
Can you manipulate an image to make it more creative?
Our paper "Exploring Latent Dimensions of Crowd-sourced Creativity" is just accepted to NeurIPS Creativity and Design workshop! Coming on Arxiv soon.
Congrats to Umut, Alperen and Efehan!
@howtogaa
@Bogazici_CmpE
Each project features a GAN or a text-generated design, and made in real life by human collaborators! Batuhan
@iambatukoyuncu
worked on a GAN-based abstract art project and Kubilay () made one of the designs in real life!
@howtogaa
@Bogazici_CmpE
@iambatukoyuncu
Giray Eryilmaz worked on a GAN-based carpet generation project. One of the designs made in real life by Mis Carpets () in Grand Bazaar (one of the oldest marketplaces in the world, and famous for its carpet artisans). The result was extraordinary!
Happy to announce that our paper "Controlled Cue Generation for Play Scripts" has been selected as the best paper at Controllable Generative Modeling in Language and Vision Workshop at NeurIPS 2021! Congrats Alara and Hilal! 🎉
Btw, these are the AI-generated graffiti made in real life by artists IMAGINE and SOBEK for
@howtogaa
() Hopefully, more artists are going to do amazing collaborations with AI with new
#styleGAN
graffiti designs!
#gan
#styleGAN
#graffiti
#art
“I love that I get to use both my creative and logical skills to help make fashion more inclusive, and by extension the tech industry as well!” .
Dominique is a singer, songwriter, model, software engineer and all-around superstar! ⭐️
#girlswhocode
#ai
#fashiontech
Let's RAVE! A fast zero-shot video editing method capable of style, attribute and shape editing, without additional training or finetuning! .
A collab between VT, Georgia Tech, UIUC w/amazing team
@ozgurkara99
@bariskurtkaya
@d_yesiltepe
@RehgJim
.
RAVE has been featured in Hugging Face's Spaces of the Week! 🔥 Welcome to try it out
@huggingface
: For more information, visit our project page: . Thanks
@_akhaliq
for the support!
If I had conformed to the expectations of others, none of this would have happened. Life is short. If you genuinely take an interest in something, follow it. You never know when those skills will come in handy! Even if you don't end up using them, you'll still enjoy the journey!
Shout out to my amazing friend Agnes Cameron, who inspired my interest in fashion during our time at MIT! She created these incredible pieces inspired by AI-generated designs (reminder: this was in 2018, so the generated images were potato quality!)
Asked to generate a heavy metal song about a generative AI lab named Gem Lab at Virginia Tech. Well, it did not disappoint🤘Gem Lab is the name, where students come to shine!
I used my daughter's birthday 132019 as the seed and this is one of the people AI generated (inspired by this number). It seems like they could be playmates! (my daughter on the right for reference)